[gmx-users] question about the implicit walls

Berk Hess gmx3 at hotmail.com
Mon Sep 21 15:20:46 CEST 2009


For all interactions except for the PME mesh part there is no pbc in z.
For PME mesh there is periodicity in z.

For c) there are several options.
One issue is that the DD setup is not very smart (yet)
and it might decompose is z. If this is case, use the -dd option
of mdrun to preferable decompose in x and possibly also in y.

Another option would be to use a slightly longer cut-off and a coarser grid.
Or simply do not use separate PME nodes.


Date: Mon, 21 Sep 2009 15:09:13 +0200
From: qiaobf at gmail.com
To: gmx-users at gromacs.org
Subject: Re: [gmx-users] question about the implicit walls


HI Berk,

Thanks for your reply!

Questions related the implicit Walls. 

a) When I use PBC =xyz in combination with nwall=2, I get the following

           ERROR: walls only work with pbc=xy

     But If I use pbc=xy, there is no periodic image in the Z
direction, why should we use "wall_ewald_zfac"=2~3? 

c) Even though I have used half of the processors for -npme, there is
still more than 20% is lost due to PME have more work to do. Is there
any idea to improve the performance?

The following is the parameters for the walls I try to simulate the
graphine substrate.


; Number of walls, type, atom types, densities and box-z scale factor
for Ewald

nwall                         = 2

wall_type                 = 9-3

wall_r_linpot           = 0

wall_atomtype        = opls_135   opls_135

wall_density            = 114  114

wall_ewald_zfac     = 3

Berk Hess wrote:


Nearly all the information is in the mdp options page, but it is a bit concise.
The only really missing information is that there is not cut-off for the wall.
All atoms feel the force of one or both walls.

The density is required to determine the interaction strength of the wall
for the 9-3 and 10-4 options. The LJ potential is integrated over the volume
behind the wall or the plain of the wall, which requires a LJ particle density.


Date: Mon, 21 Sep 2009 10:02:29 +0200
From: qiaobf at gmail.com
To: gmx-users at gromacs.org
Subject: [gmx-users] question about the implicit walls



Is anyone who can clear the "Walls" feature in the gmx 4.0.x? Same as
Yves (see the end), I cannot find any related reference or publication
citing it. 

Generally the potential on wall is in continuum format. But if it is
also true here,  what the "wall_density" for?  I guess it is based on a
lattice model. If I want to build silicon or graphene walls, what kind
of values is suitable for them? Any reference?


Baofu Qiao

[gmx-users] Implicit walls.
Yves Dubief

Thu, 11 Jun 2009 03:32:11 -0700


I am using the implicit wall method available in 4.0.x and I cannot
seem to find any published work on the method, aside from the
manual. I think I have an overall understanding of the
method, but I am not confident about some key details:
- Is the force generated from the wall derived from a
continuum representation of the wall based on the surface
density, or from a lattice of virtual atoms? I am leaning
toward #1 just because it works great under NPT
- What is the depth of the virtual wall when wall_r_linpot=0?
Is it LJ cutoff? or "zero"?
- Are any of developer willing to validate the paragraph we
plan on writing in my student thesis and future publications
(when it's all ready)?

   Thank you in advance.
Express yourself instantly with MSN Messenger! Download today it's FREE!
gmx-users mailing list    gmx-users at gromacs.org
Please search the archive at http://www.gromacs.org/search before posting!
Please don't post (un)subscribe requests to the list. Use the 
www interface or send it to gmx-users-request at gromacs.org.
Can't post? Read http://www.gromacs.org/mailing_lists/users.php

What can you do with the new Windows Live? Find out
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20090921/d16780fa/attachment.html>

More information about the gromacs.org_gmx-users mailing list